Google Alleged Outsider Enterprise Used Gemini to Create Phishing Sites for Text-Message Scams

Google alleged in a civil lawsuit that the China-based network Outsider Enterprise used Gemini and other AI tools to help create phishing websites for large-scale text-message scams impersonating companies and government services. Google said the network coordinated through Telegram, generated thousands of fake sites and fraudulent URLs, sent 2.5 million Android messages during a two-week period in May 2026, and financially scammed hundreds of thousands of people.

Risk classification

• Primary risk domain: 4 Malicious actors
• Primary risk subdomain: 4.3 Fraud, scams, and targeted manipulation

The cybercrime network used Gemini to create fake websites and mass-produce scam messages to defraud hundreds of thousands of victims for financial gain.

Additional risk subdomains

• 3.1 False or misleading information: The AI was used to generate highly convincing fake websites and deceptive messages that misled users into believing they were interacting with legitimate organizations.

Causal factors

• Entity: Human
• Intent: Intentional
• Timing: Post-deployment

The risk was caused by the intentional actions of a human cybercrime network using a deployed AI system to conduct financial scams.

EU AI Act risk tier

• Risk tier: 3 Limited Risk

Limited Risk: The AI system involved, Gemini, is a general-purpose chatbot and generative AI system used to generate text and website content, which falls under the category of chatbots and AI-generated content requiring transparency.

AI system and alleged parties

• AI system: Gemini (Google)
• AI purpose: Writing Assistant; Technical Text Generation
• Behaviour type: Assistant
• Alleged developer: Large language model developers, Google DeepMind, Google
• Alleged deployer: Scammers in China, Scammers, Outsider Enterprise, Cybercriminals in China, Cybercriminals
• Alleged harmed parties: YouTube, United States Postal Service, Google users, Google, E-ZPass, Consumers, Android users

Harm severity

Severity ratings are based on this scale.

Highest direct severity in any category: Severe. Severity is scored from Negligible to Catastrophic in each harm category, for harm the reports describe as caused directly or indirectly by the AI system.

• Physical: direct Negligible, indirect Negligible
• Infrastructure: direct Negligible, indirect Negligible
• Property: direct Negligible, indirect Negligible
• Financial: direct Substantial, indirect Negligible
• Environmental: direct Negligible, indirect Negligible
• Malicious content: direct Severe, indirect Negligible
• Differential treatment: direct Negligible, indirect Negligible
• Civil rights: direct Negligible, indirect Negligible
• Democracy: direct Negligible, indirect Negligible
• Privacy: direct Negligible, indirect Negligible
• Psychological: direct Negligible, indirect Negligible
• Epistemic: direct Minor, indirect Negligible
• Child sexual exploitation and abuse: direct Negligible, indirect Negligible

Financial

Reported: Yes, the report explicitly describes financial losses in the millions of dollars resulting from the scam campaign.

Directly caused: The report states that the financial damages are 'in the millions' of dollars. For the 1 occurrence of this coordinated campaign, the average financial loss is estimated to be in the millions (e.g., $5,000,000).

Inferred additional harm: Given that the FBI reported $893 million in AI-linked losses last year, and this specific campaign affected hundreds of thousands of people, the actual financial losses could be much higher than the initial 'millions' reported, potentially tens of millions of dollars.

Malicious content

Reported: Yes, the report describes the creation and distribution of millions of malicious phishing messages and thousands of fake websites.

Directly caused: The Chinese network used Gemini to create 131 software kits, 9,000 fake websites, and sent 2.5 million scam messages to Android users.

Epistemic

Reported: Yes, the report describes the creation of hundreds of fake websites mimicking real companies and government operations to deceive users.

Directly caused: The network created fake websites mimicking Google, YouTube, the Postal Service, and E-ZPass, misleading users into believing they were interacting with legitimate entities.

Inferred additional harm: The widespread deployment of highly convincing fake websites and messages erodes public trust in digital communications and legitimate brand/government websites.

People affected

• Occurrences reported: 1
• People reportedly harmed: 200000
• People reportedly exposed: 2500000

Potential causes

Management

• Delayed Industry Coordination: First coordinated effort with FBI and carriers occurred after the event.

Technology

• Gemini Vulnerability to Misuse: Gemini was used to generate fake websites mimicking trusted brands.
• AI-Powered Software Kits: Criminals created 131 kits to easily mass-produce scam websites.
• High Conviction Generation: AI-generated scams are highly convincing and hard to detect.

Human Factors

• User Trust in Mimicked Brands: Victims trusted fake websites mimicking USPS, E-ZPass, and Google.
• Malicious Collaboration: Scammers used Telegram to share AI tools and coordinate campaigns.

Process and Methods

• Mass Messaging Distribution: Scammers sent 2.5 million messages to Android users in two weeks.
• Inadequate Abuse Prevention: Google lacked proactive controls to block Gemini-assisted scams.

Regulatory Environment

• Lack of Preemptive Regulation: No regulatory standard prevented the deployment of scam-enabling AI.

Information quality

• Classification confidence: High
• Reason for confidence: The reports provide consistent and detailed coverage of the lawsuit filed by Google against the cybercrime network, including specific metrics such as the number of messages sent (2.5 million), fake websites created (9,000), and software kits (131). The role of Gemini is explicitly stated, and the financial and human impact is well-scoped, leaving little ambiguity about the nature of the incident.
• Ambiguities identified: The exact amount of financial damages is described generally as 'in the millions' rather than a precise figure.
• Alternative interpretations: None.

A China-based cybercrime network named Outsider Enterprise leveraged Google's Gemini AI to scale a massive phishing campaign targeting U.S. citizens, resulting in millions of dollars in losses. The incident highlights the growing threat of AI-enabled economic fraud and the dual-use risks of large language models, prompting coordinated intervention by the FBI, Google, and major telecom carriers.

• Overall national security impact: Substantial
• Scope: Single nation
• Primary target: United States
• Alleged perpetrator: Outsider Enterprise

Threat characteristics

• Imminence: Long-term. Represents an ongoing strategic concern regarding the ease of scaling cyber fraud using commercial AI systems.
• Autonomy: Human-controlled. The AI system acted as a generation assistant, with the human cybercriminals making all final decisions and executing the campaigns.
• Novelty: Evolved capability. Represents an evolved capability where generative AI is utilized to rapidly scale and increase the persuasiveness of established phishing and fraud techniques.

Impact by dimension

• Physical security: Negligible. No physical damage, kinetic attacks, or manipulation of critical infrastructure control systems were reported.
• Information security: Minor. The campaign used AI-generated content to mimic government and corporate entities, but lacked evidence of state-sponsored intelligence compromise or strategic information warfare.
• Sovereignty: Minor. The network mimicked government services like the U.S. Postal Service and E-ZPass, but did not disrupt core government operations, decision-making, or electoral systems.
• Economic security: Substantial. China-based cybercrime network caused millions of dollars in losses to hundreds of thousands of U.S. citizens, prompting coordination between Google, the FBI, and major telecom providers.
• Societal stability: Minor. Widespread financial fraud affected hundreds of thousands of victims, eroding public trust in digital communications, but did not involve mass surveillance or systematic civil rights violations.